Principles of Survey Research (questionStar)

Paul Marx | Principles of survey research
Principles of Survey Research
1
introductory course

Contents
1. Introduction
1.1 Market Research and Survey
1.2 Types of Market Research
2. Survey: Measurement and Scaling
2.1 Introduction
2.2 Comparative Scales
2.3 Non-Comparative Scales
2.4 Multi-item Scales
2.5 Reliability and Validity
3. Questionnaire
3.1 Asking Questions
3.2 Overcoming Inability to Answer
3.3 Overcoming Unwillingness to Answer
3.4 Increasing Willingness of Respondents
3.5 Determining the Order of Questions
3.6 What’s Next?
4. Sampling
4.1 Non-probability Sampling
4.2 Probability Sampling
4.3 Choosing Non-probability vs. Probability Sampling
4.4 Sample Size
5. Data Analysis:
A Concise Overview of Statistical Techniques
5.1 Descriptive Statistics:
Some popular Displays of Data
5.1.1 Organizing Qualitative Data
5.1.2 Organizing Quantitative Data
5.1.3 Summarizing Data Numerically
5.1.4 Cross-Tabulations
5.2 Inferential Statistics:
Can the Results Be Generalized to Population?
5.2.1 Hypotheses Testing
5.2.2 Strength of a Relationship in Cross-Tabulation
5.2.3 Describing the Relationship between Two (Ratio Scaled) Variables
6. Advanced Techniques of Market Analysis:
A Brief Overview of Some Useful Concepts
6.1 Conjoint-Analysis
6.2 Market Simulations
6.3 Segmentation
6.4 Perceptual Positioning Maps
7. Reporting Results
2

1.Introduction
3

1.Introduction
4

What is Research?
Research is
the systematic investigation into and study of
materials and sources in order to establish facts
and reach new conclusions.
(Oxford Dictionaries)
5
Research is
the searching for and gathering of information
and ideas in response to a specific question.
(Unknown author)

Survey Research
6
Survey -
The most popular technique for gathering
primary data in which a researcher interacts with
people to obtain facts, opinions, and attitudes.

The Essence of Market Research
7
Researcher
Decision Maker
Obvious Measurable
Symptoms
Real Business/Decision
Problems
Unhappy
Customers
Decreased
Market
Share
Loss
of
Sales
Low
Traffic
Low-Quality
Products
Poor Image
Marginal
Performance
of Sales Force
Inappropriate
Delivery System
Unethical Treatment
of Customers
Decision Problem Definition

Who Why
Sociology and Political Science Public opinion research, identification of population's attitudes towards socially important phenomena, events, and facts…
Psychology
Personality tests, intelligence tests, identification of individual strengths and weaknesses psychological stability, cognitive
disorders, social influence…
Human Resources
Measurement of employee satisfaction, loyalty, potential, personality traits and leadership skills, productivity and quality of
work, professional fit, resistance to stress, social intelligence, work-life balance…
Marketing
Market and consumer research, measurement of perception of image, preferences, attitudes, satisfaction with product
and/or service, loyalty, willingness to pay; segmentation, positioning, new product development, evaluation of market
potentials, pricing and price setting, advertising tests, ease of web-site navigation, user feedback, willingness to
recommend...
Science (in general)
Study of relationships between two or more variables, factors, phenomena; development of scales and survey techniques
for practical use…
Education Knowledge tests (quizzes, exams), evaluation of students and/or teachers…
… …
Practical Application of Surveys
8

Market Research Process
Define the
Research problem
Develop the
research plan
Collect
data
Analyze
data
Report
findings
9
⁻ identify and clarify
information needs
⁻ define research
problem and
questions
⁻ specify research
objectives
⁻ confirm information
value
If a problem is vaguely
defined, the results can
have little bearing on the
key issues
Decide on
⁻ budget
⁻ data sources
⁻ research approaches
⁻ sampling plan
⁻ contact methods
⁻ methods of data
analysis
The plan needs to be
decided upfront but
flexible enough to
incorporate changes or
iterations
⁻ collect data according
to the plan or
⁻ employ an external
firm
This phase is the most
costly and the most liable
to error
Analyze data
⁻ statistically or
⁻ subjectively
and infer answers and
implications
Type of data analysis
depends on type of
research
- Formulate
conclusions and
implications from
data analysis
- prepare finalized
research report
Overall conclusions to be
presented rather than
overwhelming statistical
methodologies

When NOT to Conduct Market Research
Occasion Comments
Vague objectives
When managers cannot agree on what they need to know to make a decision. Market research cannot be helpful unless it is
probing a particular issue.
Closed mindset When decision has already been made. Research is used only as a rubber stamp of a preconceived idea.
Late timing When research results come too late to influence the decision.
Poor timing If a product is in a “decline” phase there is little point in researching new product varieties.
Lack of resources
If quantitative research is needed, it is not worth doing unless a statistically significant sample can be used. When funds are
insufficient to implement any decisions resulting from the research.
Costs outweigh benefits The expected value of information should outweigh the costs of gathering an analyzing the data..
Results not actionable Where, e.g., psychographic data is used which will not help he company form firm decisions.
10

1.Introduction
11

Types of Market Research
12
By Objectives
• Exploratory
(a.k.a. diagnostic)
• Descriptive
• Causal
(a.k.a. predictive, experimental)
By Data Source
• Primary
• Secondary
By Methodology
• Qualitative
• Quantitative

Market Research by Objectives
•Explaining data or actions to help define the problem
•What was the impact on sales after change in the package design?
•Do promotions at POS influence brand awareness?
Exploratory
a.k.a. diagnostic
•Gathering and presenting factual statements:
who, what, when, where, how
•What is historic sales trend in the industry?
•What are consumer attitudes toward our product?
Descriptive
•Probing cause-and-effect relationships; “What if?”
•Specification of how to use the research to predict
•the results of planned marketing decisions
•Does level of advertising determine level of sales?
Causal
a.k.a. predictive, experimental
13
Survey
of a small
sample, focus
groups, depth
interviews,,…
Survey
of a large
representative
sample,
observation, …
Experiments,
A&B tests,
consumer
panels, …
Uncertainty influences the type of research
UncertainCertain

Market Research by Data Source
14
• Original research to collect new raw data for a specific
reason. This data is then analyzed and may be published
by the researcher.
Primary
• Research data that has been previously collected,
analyzed and published in the form of books, articles,
etc.
Secondary
Survey,
Interviews,
observation,
experiments, …
Literature
review, library,
web, database,
archive,…

Market Research by Methodology
15
• Involves collecting and measuring data
• Often requires large data sets. For example, large number of people.
• Uses statistical methods to analyze data
• Aims to achieve objective/scientific view of the subject
Quantitative
• Involves understanding human behavior and the reasons behind it
• Focus is on individuals and small groups
• Objectivity is not the goal, the aim is to understand one point of view, not all
points of view.
• Usually not representative
Qualitative
Survey
of a large
representative
sample,
observation, …
Survey
of a small
sample, focus
groups, depth
interviews,,…

Paul Marx | Principles of survey research 16
APPARENT
TRUTH
Literature Review
InterviewSurvey
Triangulation
Robson (1998), Visocky & Visocky (2009)

2.Survey: Measurement and Scaling
2.1 Introduction
18

2.1 Introduction
19

Measurement
Measurement –
assigning numbers or other symbols to
characteristics of objects according to certain pre-
specified rule
- one-to-one correspondence between the
numbers and characteristics being
measured
- the rules for assigning numbers should be
standardized and applied uniformly
- rules must not change over objects or time
20

Scaling
Scaling –
involves creating a continuum upon
which measured objects are located.
21
Extremely
favorable
Extremely
unfavorable

Primary Scales of Measurement
22
• numbers serve as labels for identifying and classifying objects
• not continuosNominal
• numbers indicate the relative positions of objects
• but not the magnitude of difference between themOrdinal
• differences between objects can be compared
• zero point is arbitraryInterval
• zero point is fixed
• ratios of scale values can be computedRatio
a.k.a. metric
or
1 2 1 2 1 2
NOT
3
1
2
1 2 3
My preference as a snack food
moreless
0 25 50 75 100
Weight(kg)

Primary Scales of Measurement
Scale Basic Characteristics
Common
Examples
Marketing
Examples
Permissible Statistics
Descriptive Inferential
Nominal Numbers identify and classify
objects
Social security
numbers, numbering
of football players
Brand numbers, store
types sex,
classification
Percentages, mode Chi-square,
binomial test
Ordinal Numbers indicate the relative
positions of the objects but
not the magnitude of
differences between them
Quality rankings,
ranking of teams in
tournament
Preference rankings,
market position, social
class
Percentile, median Rank-order
correlation,
Friedman ANOVA
Interval Differences between objects
can be compared; zero point
is arbitrary
Temperature
(Fahrenheit,
Centigrade)
Attitudes, opinions,
index numbers
Range, mean,
standard deviation
Product-moment
correlations, t-tests,
ANOVA, regression,
factor analysis
Ratio Zero point is fixed; ratios of
scale values can be
compared
Length, weight,
time, money
Age, income, costs,
sales, market shares
Geometric mean,
harmonic mean
Coefficient of
variation
23

Classification of Scaling Techniques
Scaling
Techniques
Comparative
Scales
Paired
Comparison
Rank Order Constant Sum
Q-Sort &
others
Non-
comparative
Scales
Continuous
Rating Scales
Itemized
Rating Scales
Likert
Semantic
Differential
Stapel
24

Comparison of Scaling Techniques
25
Comparative
Scales
• involve the direct comparison
of stimulus objects.
• data must be interpreted in
relative terms
• have only ordinal and rank-
order properties
Non-comparative
Scales
• each object is scaled
independently
• resulting data is generally
assumed to be interval or
ratio scaled
- nature of the research
- variability in the population
- statistical considerations

2.1 Introduction
26

Scaling
Techniques
Comparative
Scales
Paired
Comparison
Q-Sort &
others
Non-
comparative
Scales
Continuous
Rating Scales
Itemized
Rating Scales
Likert
Semantic
Differential
Stapel
27

Relative Advantages of Comparative Scales
28
+ small differences between stimulus
objects can be detected
+ same known reference points for all
respondents
+ easy to understand and to use
+ involve fewer theoretical assumptions
+ tend to reduce halo or carryover
effects from one judgement to another
Advantages
- have only ordinal and rank-order
properties ⟶ limited set of statistical
methods available for analysis
- data must be interpreted in relative
terms
- Inability to generalize beyond the set
of compared objects
Disadvantages

Comparative Scales: Paired Comparison
29
Respondent is presented with two objects
and asked to select one according to
some criterion
We are going to present you with ten pairs of beer brands. For
each pair, please indicate which one of the two brands of beer
you would prefer to purchase.
Heineken Beck’s Coors Budweiser Miller
Heineken
Beck’s
Coors
Budweiser
Miller
#Preferred 3 2 0 4 1
Paired Comparison

Paired Comparison Scales: Examples
30

Paired Comparison: Pros-and-Cons
31
+ direct comparison and overt choice
+ good for blind tests, physical products, and
MDS
+ allows for calculation of percentage of
respondents who prefer one stimulus to
another
+ can assess rank-orders of stimuli (under
assumption of transitivity)
+ possible extensions: “no difference”
alternative; graded comparison
Advantages
- # of comparisons grows quicker than # of
stimuli (for n objects n(n-1)/2 comparisons)
- presentation order bias possible
- preference of A over B does not imply
subject’s liking of A
- little similarity to real choice situation with
multiple alternatives
- violations of transitivity may occur
Disadvantages

>
>
Ordinal Data:
violations of transitivity in paired comparison
32

Ordinal data:
violations of transitivity when aggregating preferences
33
Respondent #1
Respondent #2
Respondent #3
Votes count
Result:
2 vs 1
2 vs 1
2 vs 1
Apple is both the best and the worst alternative.
Aggregated preferences of the group are inconsistent!
Voting

Comparative Scales: Rank Order Scaling
34
Respondents are presented with several
objects simultaneously and are asked to
order or rank them according to some
criterion
Rank the various brands of soft drinks in order of preference.
Begin by picking out the one brand that you like most and
assign it a number 1. Then find the second most preferred brand
and assign it a number 2. Continue this procedure until you have
ranked all the brands of soft drinks in order of preference. The
least preferred brand should be assigned a rank of 5.
No two brands should receive the same rank number.
The criterion of preference is entirely up to you. There is no right
or wrong answer. Just try to be consistent.
Rank Order Scaling
Brand Rank Order
Pepsi ______________
Coke ______________
Red Bull ______________
Mountain Dew ______________
Kvas ______________

Rank Oder Scales: Example
35
©ExavoGmbH, exavo.de

Rank Oder Scales: Examples
36

Rank Oder Scales: Example
37

Rank Oder Scales: Pros-and-Cons
38
+ direct comparison
+ more realistic than paired comparison
+ # of comparisons is only (n-1)
+ easier to understand
+ takes less time
+ no intransitive responses
+ can be converted to paired
comparison data
+ good for measuring preferences of brands
or attributes; conjoint analysis
Advantages
- preference of A over B does not imply
subject’s liking of A
- no zero point / separation between liking
and disliking
- only ordinal data
- violations of transitivity may occur
Disadvantages

Comparative Scales: Constant Sum Scaling
39
Respondents allocate a constant sum of
units (points, dollars, chips, %) among a
set of stimulus objects with respect to
some criterion
Below are five attributes of cars. Please allocate 100 points
among the attributes so that your allocation reflects the relative
importance you attach to each attribute. The more points an
attribute receives, the more important the attribute is. If an
attribute is not at all important, assign it zero points. If an
attribute is twice as important as some other attribute, it should
receive twice as many points.
Constant Sum
Attribute Points
Speed 0
Comfort 15
Gear Type 5
Fuel Type
(gasoline/diesel)
35
Price 45
sum 100

Constant Sum Scaling: Example of Analysis
40
Attribute Segment 1 Segment 2 Segment 3
Speed 0 17 53
Comfort 15 23 30
Gear Type 5 21 10
Fuel Type
(gasoline/diesel)
35 12 7
Price 45 27 0
sum 100 100 100
Average response of three segments

Constant Sum Scaling: Example
41

Constant Sum Scaling: Examples
42

Constant Sum Scaling: Pros-and-Cons
43
+ allows for fine discrimination among
stimulus objects without requiring too
much time
+ ratio scaled ⟶ flexible choice of data
analysis methods
Advantages
- results are limited to the context of stimuli
scaled, i.e., not generalizable to other
stimuli not included in the study
- relatively high cognitive burden for
respondents, esp. when # of items is large
- prone to calc. errors (e.g., allocation
of 108 or 94 points)
Disadvantages

Comparative Scales: Q-Sort Scaling
44
A rank order procedure in which objects are
sorted into piles based on similarity with
respect to some criterion. Usually used to
discriminate among a relatively large number
(60-140) of objects quickly.
The number of objects in each pile is limited,
usually so that all piles imitate normal
distribution.
To prevent epidemics, the Ministry of Health has developed the
following 25 measures recommended for implementation in
hospitals. Please distribute these measures for preventing the
spread of infections according to their importance using the
scheme below. Please allocate only one measure per box. Q-Sort
not at all
important
extremely
important

2.1 Introduction
45

Scaling
Techniques
Comparative
Scales
Paired
Comparison
Q-Sort &
others
Non-
comparative
Scales
Continuous
Rating Scales
Itemized
Rating Scales
Likert
Semantic
Differential
Stapel
46

Non-Comparative Scales: Continuous Rating Scale
47
Respondents rate objects by placing a mark at
the appropriate position on a line that runs
from one extreme of the criterion variable to
the other.
How would you rate Wal-Mart as a department store?
Continuous Rating Scale
Probably
the worst
Probably
the best
Version 1
х
Probably
the worst
Probably
the best
Version 2
х0 10 20 30 40 50 60 70 80 90 100
Probably
the worst
Probably
the best
Version 3
х0 20 40 60 80 100
very bad very good
neither good
nor bad
Probably
the worst
Probably
the best
Version 4
very bad very good
neither good
nor bad
76

Continuous Rating Scale: Perception Analyzer
48

Itemized Rating Scales: Likert Scale
49
Requires respondents to indicate a degree of
agreement or disagreement with each of a
series of statements about the stimulus object
within typically five to seven response
categories.
Listed below are different opinions about 7-Eleven. Please
indicate how strongly you agree or disagree with each by using
the following scale:
Likert Scale Strongly
disagree
Disagree Neither agree
nor disagree
Agree Strongly
agree
7-Eleven sells high-quality
merchandise
[1] [x] [3] [4] [5]
7-Eleven has poor in-store
service
[1] [x] [3] [4] [5]
I like to shop in 7-Eleven [1] [2] [x] [4] [5]
7-Eleven does not offer a
good mix of different brands
within a product category
[1] [2] [3] [x] [5]
The credit policies at 7-Eleven
are terrible
[1] [2] [3] [x] [5]
I do not like advertising done
by 7-Eleven
[1] [2] [3] [x] [5]
7-Eleven charges fair prices [1] [x] [3] [4] [5]
NOTICE the reversed scoring of items 2,4,5, and 6. Reverse the scale for these items prior analyzing
to be consistent with the whole set of items, i.e. a higher score should denote a more favorable attitude.

Likert Scale: Examples
50

Some Commonly Used Scales in Marketing
51
Construct Scale Descriptors
Attitude Very bad Bad Neither Bad Nor
Good
Good Very Good
Importance Not at All
Important
Not Important Neutral Important Very
Important
Satisfaction Very Dissatisfied (Somewhat)
Dissatisfied
Neither Dissatisfied
Nor Satisfied /
Neutral
(Somewhat)
Satisfied
Very Satisfied
Purchase Intention Definitely Will Not
Buy
Probably will Not
Buy
Might or Might Not
Buy
Probably Will
Buy
Definitely Will
Buy
Purchase Frequency Never Rarely Sometimes Often Very Often
Agreement Strongly Disagree Disagree Neither Agree Nor
Disagree
Agree Strongly Agree

Itemized Rating Scales: Semantic Differential
52
A rating scale with end point associated with
bipolar labels that have semantic meaning.
Respondents are to indicate how accurately or
inaccurately each term describes the object.
This part of the study measures what certain department stores
mean to you by having you judge them on a series of descriptive
scales bounded at each end by one of two bipolar adjectives.
Please mark (X) the blank that best indicates how accurately
one or the other adjective describes what the store means to
you. Please be sure to mark every scale; do not omit any scale.Semantic Differential
Powerful [ ] [ ] [ ] [ ] [X] [ ] [ ] Weak
Unreliable [ ] [ ] [ ] [ ] [ ] [X] [ ] Reliable
Modern [ ] [ ] [ ] [ ] [ ] [ ] [X] Old fashioned
Cold [ ] [ ] [ ] [ ] [ ] [X] [ ] Warm
Careful [ ] [X] [ ] [ ] [ ] [ ] [ ] Careless
NOTE: The negative adjective sometimes appears at the left side of the scale and sometimes
at the right. This controls the tendency of some respondents, particularly those with very positive
or very negative attitudes, to mark the right- or left-hand sides without reading the labels.
7-Eleven is:

Semantic Differential Scale: Example
53
Rugged [ ] [ ] [ ] [ ] [ ] [ ] [ ] Delicate
Excitable [ ] [ ] [ ] [ ] [ ] [ ] [ ] Calm
Uncomfortable [ ] [ ] [ ] [ ] [ ] [ ] [ ] Comfortable
Dominating [ ] [ ] [ ] [ ] [ ] [ ] [ ] Submissive
Thrifty [ ] [ ] [ ] [ ] [ ] [ ] [ ] Indulgent
Pleasant [ ] [ ] [ ] [ ] [ ] [ ] [ ] Unpleasant
Contemporary [ ] [ ] [ ] [ ] [ ] [ ] [ ] Non-contemporary
Organized [ ] [ ] [ ] [ ] [ ] [ ] [ ] Unorganized
Rational [ ] [ ] [ ] [ ] [ ] [ ] [ ] Emotional
Youthful [ ] [ ] [ ] [ ] [ ] [ ] [ ] Mature
Formal [ ] [ ] [ ] [ ] [ ] [ ] [ ] Informal
Orthodox [ ] [ ] [ ] [ ] [ ] [ ] [ ] Liberal
Complex [ ] [ ] [ ] [ ] [ ] [ ] [ ] Simple
Colorless [ ] [ ] [ ] [ ] [ ] [ ] [ ] Colorful
Modest [ ] [ ] [ ] [ ] [ ] [ ] [ ] Vain
Measuring Self-Concepts, Person Concepts,
and Product Concepts
Rating profiles of different objects / respondents / segments.
Each point corresponds to a mean or median of the respective scale.

54
Source: http://www.provisor.com.ua/archive/2000/N16/gromovik.php
Cheap [ ] [ ] [ ] [ ] [ ] [ ] [ ] Expensive
Has natural ingredients [ ] [ ] [ ] [ ] [ ] [ ] [ ]
Has no natural
ingredients
Attractive [ ] [ ] [ ] [ ] [ ] [ ] [ ] Unattractive
Easily available [ ] [ ] [ ] [ ] [ ] [ ] [ ] Hard to get
Smells good [ ] [ ] [ ] [ ] [ ] [ ] [ ] Smells bad
Has conditioner [ ] [ ] [ ] [ ] [ ] [ ] [ ] Has no conditioner
Well-known brand [ ] [ ] [ ] [ ] [ ] [ ] [ ] Unknown brand
Suitable for
frequent usage
[ ] [ ] [ ] [ ] [ ] [ ] [ ]
Unsuitable for
frequent usage
Miraculous effect of
cleanliness and shine
[ ] [ ] [ ] [ ] [ ] [ ] [ ]
Lack of cleanliness
effect
Easy-to-use [ ] [ ] [ ] [ ] [ ] [ ] [ ] Inconvenient to use
Ideal shampoo
Elseve
Herbal Magic
Semantic profiles of shampoo brands
“Herbal Magic” and “Elseve” in comparison with
an ideal shampoo from consumers’ point of view

55

Itemized Rating Scales: Stapel Scale
56
An unipolar rating scale with 10 categories
numbered from -5 to +5 without neutral point
(zero).
Used as an alternative to semantic differential,
especially when a meaningful pair of opposed
adjectives is difficult to construct.
Please evaluate how accurately each word or phrase describes each
of department stores. Select a plus number for phrases you think
describe the store accurately. The more accurately you think the
phrase describes the store, the larger the plus number you should
choose. You should select a minus number for phrases you think do
not describe it accurately. The less accurately you think the phrase
describes the store, the larger the minus number you should choose.
You can select any number, from +5 for phrases you think are very
accurate, to -5 for phrases you think are very inaccurate.
Stapel Scale
7-Eleven:
+5
+4
+3
+2
+1
-1
-2
-3
-4
-5
High Quality
+5
+4
+3
+2
+1
-1
-2
-3
-4
-5
Poor service
х
х

Basic Non-Comparative Scales
Scale Basic Characteristics Examples Advantages Disadvantages
Continuous Rating
Scale
Place a mark on a continuous line Reaction to TV
commercials
Easy to construct Scoring can be
cumbersome, unless
computerized
Itemized Scales
Likert  Scale Degrees of agreements on a 1
(strongly disagree) to 5 (strongly
agree) scale
Measurement of
attitudes
Easy to construct, administer
and understand
More time-consuming
Semantic
Differential
Seven-point scale with bipolar labels Brand, product, and
company images
Versatile Controversy as to whether
the data are interval
Stapel
Scale
Unipolar ten-point scale, -5 to +5,
without a neutral point (zero)
Measurement of
attitudes and images
Easy to construct, administer
over telephone
Confusing an difficult to
apply
57

Non-comparative Itemized Rating Scale Decisions
58
Number of categories
Although there is no single, optimal number, traditional guidelines
suggest that there should be between five and nine categories.
Balanced vs. unbalanced In general, the scale should be balanced to obtain objective data.
Odd/even no. of categories
If a neutral or indifferent scale response is possible for at least
some respondents, an odd number of categories should be used.
Forced vs. non-forced
In situations where the respondents are expected to have no
opinion, the accuracy of the data may be improved by a non-
forced scale.
Verbal description
An argument can be made for labeling all or many scale
categories. The category descriptions should be located as close
to the response categories as possible.

Number of categories
Although there is no single, optimal number, traditional guidelines
suggest that there should be between five and nine categories.
Number of Scale Categories
59
+ The greater the number of scale
categories, the finer the discrimination
among stimulus objects that is possible
- Most respondents cannot handle more
than a few categories
Involvement and knowledge
• more categories when respondents are
interested in the scaling task or are
knowledgeable about the objects
Nature of the objects
• do objects lend themselves to fine
discrimination?
Mode of data collection
• less categories in telephone interviews
Data analysis
• less categories for aggregation, broad
generalizations or group comp.
• more categories for sophisticated statistical
analysis, esp. correlation based ones

Balanced vs. unbalanced In general, the scale should be balanced to obtain objective data.
Balanced vs. Unbalanced Scales
60
Extremely good
Very good
Neither good nor bad
Very bad
Extremely bad
Balanced Scale
Extremely good
Very good
Good
Somewhat good
Bad
Very bad
Unbalanced Scale

Odd/even no. of categories
If a neutral or indifferent scale response is possible for at least
some respondents, an odd number of categories should be used.
Odd or Even Number of Categories
61
- The middle option of an attitudinal scale
attracts a substantial # of respondents
who might be unsure about their opinion
or reluctant to disclose it
- This can distort measures of central
tendency and variance
- Do we want/need “contrast” in
controversial attitudes?

Forced vs. non-forced
In situations where the respondents are expected to have no
opinion, the accuracy of the data may be improved by a non-
forced scale.
Forced vs. Non-Forced
62
- Questions that exclude the "don't know"
option tend to produce a greater volume
of accurate data
- Are respondents unwilling to answer vs.
don’t have an opinion?
- Use "don't know" or better “not
applicable” option for factual questions,
but not for attitude questions
- Use branching to ensue concept
familiarity on the respondent’s side

Verbal description
An argument can be made for labeling all or many scale
categories. The category descriptions should be located as close
to the response categories as possible.
Verbal Description
63
- Providing a verbal description for each
category may not improve the accuracy or
reliability of the data vs. scale ambiguity
- Peaked vs. flat response distributions
completely
disagree
completely
agree
disagree agree

2.1 Introduction
64

Latent Constructs
65
Please indicate how satisfied you were with your purchase
of _____ by checking the space that best gives your
answer.
satisfied [ ] [ ] [ ] [ ] [ ] [ ] [ ] dissatisfied
pleased [ ] [ ] [ ] [ ] [ ] [ ] [ ] displeased
favorable [ ] [ ] [ ] [ ] [ ] [ ] [ ] unfavorable
pleasant [ ] [ ] [ ] [ ] [ ] [ ] [ ] unpleasant
I like it very much [ ] [ ] [ ] [ ] [ ] [ ] [ ] I didn't like it at all
contented [ ] [ ] [ ] [ ] [ ] [ ] [ ] frustrated
delighted [ ] [ ] [ ] [ ] [ ] [ ] [ ] terrible
α=0,84
A Latent Construct
is a variable that cannot be
observed or measured directly but
can be inferred from other
observable measurable variables.
Thus, the researcher must capture the
variable through questions
representing the presence/level of the
variable in question.

Latent Constructs & Multi-Item Scales
Construct Dimensions Factors Items Scale
customer
satisfaction
satisfaction
with product
satisfaction
with service
friendliness
expertise
liability
the salesperson
was appealing
the salesperson
smiled to me
the salesperson
was courteous
strongly
agree
largely
agree
largely
disagree
strongly
disagree

Advantages
+ allow to assess abstract concepts
+ make it easier to understand the data and
phenomenon
+ reduce dimensionality of data through
aggregating a large number of observable
variables in a model to represent an
underlying concept
+ link observable (“sub-symbolic”) data of the
real world to symbolic data in the modeled
world
Latent Constructs & Multi-Item Scales
67

Multi-Item Scales: Make or Steal
Generate an initial pool of items:
theory, secondary data, and qualitative research
Select a reduced set of items based on
qualitative judgement
Collect data from a large pretest sample
Perform statistical analysis
Develop a purified scale
Collect more date form a different sample
Evaluate scale reliability, validity, and
generalizability
Prepare the final scale
Develop a theory
Brunner, Gordon C. II (2012), “Marketing Scales
Handbook: A Compilation of Multi-Item Measures for
Consumer Behavior & Advertising Research”, Vol. 6,
available as PDF at www.marketingscales.com/research
Journal of the Academy of Marketing Science (JAMS)
Journal of Advertising (JA)
Journal of Consumer Research (JCR)
Journal of Marketing (JM)
Journal of Marketing Research (JMR)
Journal of Retailing (JR)

Secure Customer Index™
Assessing Consumer Loyalty and Retention
69
Secure
Customer
Very satisfied
Definitely would
recommend
Definitely will
use again
D. Randall Brandt (1996), “Secure Customer Index”, Maritz Research
Overall Satisfaction 4 = very satisfied
3 = somewhat satisfied
2 = somewhat dissatisfied
1 = very dissatisfied
Willingness to
Recommend
5 = definitely would recommend
4 = probably would recommend
3 = might or might not recommend
2= probably would not recommend
1= definitely would not recommend
Likelihood to Use
Again
5 = definitely will use again
4 = probably will use again
3= might or might not use again
2= probably will not use again
1 = definitely will not use again
Secure Customers % very satisfied/definitely would repeat/definitely would recommend
Favorable Customers % giving at least "second best" response on all three measures of satisfaction and loyalty
Vulnerable Customers % somewhat satisfied/might or might not repeat/might or might not recommend
At Risk Customers % somewhat satisfied or dissatisfied/probably or definitely would not repeat/probably or
definitely would not recommend

Extended Secure Customer Index™ Burke Inc.
70
Overall Satisfaction What is your overall level of satisfaction with (BRAND/CO)?
Willingness to Recommend If you were asked to recommend a (INDUSTRY) how likely would you be to recommend
(BRAND/CO.)?
Likelihood to Repurchase
How likely are you to continue using (BRAND/CO.)?
Earned Loyalty
(BRAND/CO.) has earned my loyalty
Preferred Company
I prefer (BRAND/CO.) to all other providers
Burke Inc. http://www.burke.com/library/whitepapers/sci_white_paper_low_res_pages.pdf
Loyalty
Index
Share of Wallet
(0% - 100%)
Period 1 Period 2

2.1 Introduction
71

Multi-Item Scales: Measurement Accuracy
72
The True Score Model
ХO = ХT + ХS + ХR
where
ХO = the observed score of measurement
ХT = the true score of characteristic
ХS = systematic error
ХR = random error

Reliability & Validity
73
Reliability
• extent to which a scale produces consistent
results in repeated measurements
• absence of random error
(ХR ⟶0 |⇒ ХO ⟶ ХT + ХS)
• reliability of a multi-item scale is denoted as
Cronbach’s alpha (0 ≥ α ≥ 1)
• values of α ≥ 0,7 are considered satisfactory
Validity
• extent to which differences in observed scale
scores reflect true differences among objects
on the characteristic being measured
• no measurement error
(ХS ⟶ 0, ХR ⟶ 0 |⇒ ХO ⟶ХT)
Reliable
Not Valid
Low Validity
Low Reliability
Not Reliable
Not Valid
Both Reliable
and Valid
* α can take on also negative values, however, they cannot be interpreted

Reliable
Not Valid
Low Validity
Low Reliability
Not Reliable
Not Valid
Both Reliable
and Valid
Relationship between Reliability & Validity
74
• validity implies reliability
(ХO = ХT |⇒ ХS = 0, ХR = 0)
• unreliability implies invalidity
(ХR ≠ 0 |⇒ ХO = ХT + ХR ≠ ХT)
• reliability does not imply validity
(ХR = 0, ХS ≠ 0 |⇒ ХO = ХT + ХS ≠ ХT)
• reliability is a necessary, but not sufficient,
condition of validity

“The purpose of a scale is to allow us to represent respondents
with the highest accuracy and reliability. We can’t have one
without the other and still believe in our data.”
Bart Gamble
vice president client services,
Burke, Inc. 2000-2003

Net Promoter Score®
competitive growth rates?
76
0 1 2 3 4 5 6 7 8 9 10
Reichheld, Fred (2003) "One Number You Need to Grow", Harvard Business Review
Detractors Passives Promoters
Net Promoter Score % Promoters % Detractors= –
How likely are you to recommend company/brand/product X
to a friend/colleague/relative?
Is the scale reliable?
Is the scale valid?
NPS (-100% – +100%)
5-10% average companies
45% high potentials with open growth opportunity
50-80% market leaders with high growth potential

Net Promoter Score®: Warning
77
“Though the “would recommend” question is far and away the
best single-question predictor of customer behavior across a
range of industries, it’s not the best for every industry…So,
companies need to do their homework. They need to validate
the empirical link between survey answers and subsequent
customer behavior for their own business.”
Fred Reichheld, 2011
Reichheld, Fred, with Rob Markey (2011). The Ultimate Question 2.0. Boston: Harvard Business Review Press; pp.50-51.
?

3.Questionnaire
3.6 What’s Next?
78

Questionnaire
79
A Questionnaire – is a formalized set of
questions for obtaining information from
respondents.
Objectives of a Questionnaire:
• translate the information need into a set of
specific questions that the respondents can
and will answer
• uplift, motivate, and encourage
respondents to become involved in the
interview, to cooperate, and to complete
the interview
• minimize response error
Questionnaire

Questioning Tactics
80
• Choose an answer form a list of answer choices
• +: easy to analyze, do not task respondents’ memory and make less stress
• –: automatic and snap answers
• Response options are not set
• +: unlimited range of possible responses, “tests” respondent’s memory
• –: complexity of coding and analysis, respondents may refuse to answer
Closed-ended
Open ended
• Do you drink alcohol every day?
• What drinks do you prefer for dinner?
Direct
Indirect

Bias in Formulation
81
Q: Do you approve smoking whilst praying?
A: No
Q: Do you approve praying whilst smoking?
A: Yes
0 15 30 45 60
Yes
No
Uncertain
Do you actually believe in the big love?
Do you believe in the big love?
Noelle-Neumann and Petersen (1998), p. 192
n = 2100,
p <.05

Issues to Consider in Questionnaire Design
82
• Is the question necessary?
• Are several questions needed instead of one?
• Is the respondent informed?
• Can the respondent remember?
• Effort required of the respondents
• Sensitivity of question
• Legitimate purpose
• Cultural issues
• Ease of completion
• Comprehensiveness
• Bias in formulation

3.Questionnaire
3.6 What’s Next?
83

Asking Questions
84
“It is not every question that deserves an answer”
Publius Syrus
roman, 1st century B.C.
• Avoid ambiguity, confusion, and
vagueness
• Avoid jargon, slang, abbreviations
• Avoid double-barreled questions
• Avoid leading
• Avoid implicit assumptions
• Avoid implicit alternatives
• Avoid treating respondent’s belief about a
hypothesis as a test of the hypothesis
• Avoid generalizations and estimates

Avoid Ambiguity, Confusion and Vagueness
85
Define the issue in terms of who, what, when,
where, why, and way (the six Ws). Who, what,
when, and where are particularly important.
• Example:
Which brand of shampoo do you use?
• Ask instead:
Which brand or brands of shampoo have you
personally used at home during the last month? In
case of more than one brand, please list all the
brands that apply.

86
The W’s Defining the Question
Who The Respondent
It is not clear whether this question relates to the individual respondent or, e.g., the
respondent’s total household
What The Brand of Shampoo
It is unclear how the respondent is to answer this question if more than one brand is used
When Unclear
The time frame is not specified in this question. The respondent could interpret it as meaning
the shampoo used this morning, this, week, or over the past year.
Where Unclear
At home, at gym, on the road?
Which brand of shampoo do you use?

87
• Example:
What brand of computer do you own?
☐ Windows
☐ Mac OS
• Ask instead:
Do you own a Windows PC? (☐ Yes ☐ No)
Do you own an Apple computer? (☐ Yes ☐ No)
• Even better:
What brand of computer do you own?
☐ Do not own a computer
☐ Windows
☐ Mac OS
☐ Other
• Example:
Are you satisfied with your current auto insurance?
☐ Yes
☐ No
• Ask instead:
Are you satisfied with your current auto insurance?
☐ Yes
☐ No
☐ Don’t have auto insurance
• Even better (branch questions):
1. Do you currently have a life insurance policy?
(☐ Yes ☐ No). If no, go to question 3.
2. Are you satisfied with your current auto insurance?
(☐ Yes ☐ No)

88
Example:
In a typical month, how often do you shop in department
stores?
☐ Never
☐ Occasionally
☐ Sometimes
☐ Often
☐ Regularly
• Ask instead:
In a typical month, how often do you shop in department
stores?
☐ Less than once
☐ 1 or 2 times
☐ 3 or 4 times
☐ More than 4 times
Whenever using words “will”, “could”, “might”, or
“may” in a question, you might suspect that the
question asks a time-related question.

Avoid Jargon, Slang, Abbreviations
89
Use ordinary words
• Example:
Do you think the distribution of soft drinks is adequate?
• Ask instead:
Do you think soft drinks are readily available when you
want to buy them?
• Example:
What was your AGI last year?
$ _______

Avoid Double-Barreled Questions
90
Are several questions needed instead of one?
• Example:
Do you think Coca-Cola is a tasty and refreshing soft
drink?
• Ask instead:
1. Do you think Coca-Cola is a tasty soft drink?
2. Do you think Coca-Cola is a refreshing soft drink?

Avoid Leading
91
If you want a certain answer - why ask?
• Example:
Do you help the environment by using canvas shopping
bags?
• Ask instead:
Do you use canvas shopping bags?

Avoid Implicit Assumptions
92
The answer should not depend on upon implicit assumptions
about what will happen as a consequence.
• Example:
Are you in favor of a balanced budget?
• Ask instead:
Are you in favor of a balanced budget if it would result in
an increase in the personal income tax?

http://www.kostenlose3dmodelle.com/
mensch-argere-dich-nicht-lightwavedice
-studio-3ds-obj-lwo/
Avoid implicit alternatives
93
An alternative that is not explicitly expressed in the options is
an implicit alternative.
• Example:
Do you like to fly when traveling short distances?
• Ask instead:
Do you like to fly when traveling short distances, or would
you rather drive?

Avoid Treating Beliefs as Real Facts
94
Beliefs are only a biased representation of reality
• Example:
Do you think more educated people wear fur clothing?
• Ask instead:
1. What is your education level?
2. Do you wear fur clothing?

Avoid Generalizations and Estimates
95
Don’t task respondents’ memory and math skills
• Example:
What is the annual per capita expenditure on groceries in
your household?
• Ask instead:
1. What is the monthly (or weekly) expenditure on
groceries in your household?
2. How many members are there in your household?

3.Questionnaire
3.6 What’s Next?
96

Overcoming Inability to Answer
97
Is the Respondent Informed?
Can the Respondent Remember?
Can the Respondent Articulate?

98
Is the Respondent Informed?
Respondents will often answer questions even though they are
not informed
• Example:
Please indicate how strongly you agree or disagree with
the following statement:
“The National Bureau of Consumer Complaints provides
an effective means for consumers who have purchased a
defective product to obtain relief”
51.9% of the lawyers and 75% of the public expressed
their opinion, although there is no such entity as the
NBCC
• Use Filter Questions:
e.g. ask about familiarity and/or frequency of patronage in
a study of 10 department stores
• Use “don’t know” Option

Can the Respondent Remember?
99
The inability to remember leads to errors of omission,
telescoping, and creation
• Example:
How many liters of soft drinks did you consume during the
last four weeks?
• Ask instead:
How often do you consume soft drinks in a typical week?
☐ Less than once a week
☐ 1 to 3 times per week
☐ 4 or 6 times per week
☐ 7 or more times per week
• Use aided recall approach (where appropriate)
“What brands of soft drinks do you remember being
advertised last night on TV?”
vs
“Which of these brands were advertised last night on TV?”

Can the Respondent Articulate?
100
If unable to articulate their responses, respondents are likely
to ignore the question and quit the survey
• Example:
If asked to describe the atmosphere of the department
store they would prefer to patronage, most respondents
may be unable to phrase their answers.
• Provide aids, e.g., pictures, maps, descriptions
If the respondents are given alternative descriptions of
store atmosphere, they will be able to indicate the one
they like the best.

3.Questionnaire
3.6 What’s Next?
101

Overcoming Unwillingness to Answer
102
Most respondents are unwilling to
• devote a lot of effort to provide information
• respond to questions that they consider to be
inappropriate for the given context
• divulge information they do not see as serving a legitimate
purpose
• disclose sensitive information

103
Minimize the effort required of respondents
• Example:
Please list all the departments from which you purchased
merchandise on your most recent shopping to a
department store.
• Ask instead:
In the list that follows, please check all the departments
from which you purchased merchandise on your most
recent shopping to a department store.
☐ Women’s dresses
☐ Men’s apparel
☐ Children’s apparel
☐ Cosmetics
…….
☐ Jewelry
☐ Other (please specify) _________________

104104
Some questions may seem appropriate in certain contexts but
not in others
• Example:
Questions about personal hygiene habits may be
appropriate when asked in a survey sponsored by the
Medical Association, but not in one sponsored by a fast-
food restaurant.
• Provide context by making a statement:
“As a fast-food restaurant, we are very concerned about
providing a clean and hygienic environment for our
customers. Therefore, we would like to ask you some
questions related to personal hygiene.”

105105105
Explain why the data is needed
• Example:
Why should a firm marketing cereals want to know the
respondents’ age, income, and occupation?
• Legitimate the information request:
“To determine how the consumption of cereals vary
among people of different ages, incomes, and occupation,
we need information on ...”

3.Questionnaire
3.6 What’s Next?
106

• Place sensitive topics at the end of the questionnaire
• Preface questions with a statement that the behavior is of
interest in common
• Ask the question using third-person technique: phrase the
question as if it referred to other people
• Hide the question in a group of other questions
• Provide response categories rather than asking for specific
figures
Increasing Willingness of Respondents
107
Sensitive Topics:
- money
- family life
- political and religious beliefs
- involvement in accidents or crimes
- …

3.Questionnaire
3.6 What’s Next?
108

Determining the Order of Questions
109
• Opening Questions
The opening questions should be interesting, simple, and
non-threatening.
• Type of Information
As a general guideline, basic information should be
obtained first, followed by classification, and, finally,
identification information.
• Difficult Questions
Difficult questions or questions which are sensitive,
embarrassing, complex, or dull, should be placed late in
the sequence.

Determining the Order of Questions
110
• Effect on Subsequent Questions (funneling)
General questions should precede the specific questions
1. What considerations are important to you in selecting a
department store?
2. In selecting a department store, how important is
convenience of location?
• Logical Order / Branching Questions
The question being branched should be placed as close as
possible to the question causing the branching.
The branching questions should be ordered so that the
respondents cannot anticipate what additional
information will be required.

Example: Flowchart of a Questionnaire
111
Introduction
Ownership of Store, Bank, and/or other
Charge Cards
Purchased products in a specific department store
during the last two months
How payment was made?
Ever purchased products in a
department store?
Store
Charge
Card
Bank
Charge
Card
Other
Charge
Card
Intention to use Store, Bank,
or Other Charge Cards
yes no
yes
no
Credit Cash
Other

3.Questionnaire
3.6 What’s Next?
112

What’s Next?
113113
Introduction
• Catch the respondents’ interest
• Explain the reasons & objectives
• Ask for their help
• Tell that their support is valuable
• Tell how much time it will last
• Emphasize the anonymity
• Incentivize
(non-monetary incentives)

What’s Next?
114114
Pretest! Pretest! Pretest!!!
• question content
• wording
• sequence
• form and layout
• question difficulty
• instructions…
• analysis procedures

Recap
115
1. Develop a flow chart of the information required based on the marketing research problem
• Once the entire sequence is laid out, the interrelationships should become clear
• Match up the actual data you would expect to collect from the questionnaire against the information needs listed in
the flow chart
• Be specific in the objective for each area of information and data. You should be able to write an objective for each
area so specifically that it guides your construction of the questions.
2. At this stage, put on your “critic’s” hat on and go back over the flowchart and ask
• Do I need to know it and know exactly what I am going to do with it? or
• It would be nice to know it but I do not have to have it

4.Sampling
4.4 Sample Size
116

The world’s most famous newspaper error
President Harry Truman against Thomas Dewey
Chicago Tribute prepared an incorrect headline without first
getting accurate information
Reason?
• bias
• inaccurate opinion polls

Sampling
118
Most research cannot test everyone. Instead a sample of the
whole population is selected and tested.
If this is done well, the results can be applied to the whole
population.
This selection and testing of a sample is called sampling.
If a sample is poorly chosen, all the data may be useless.
Population
the group of people we wish to
understand. Populations are often
segmented by demographic or
psychographic features (age, gender,
interests, lifestyles, etc.)
Sample
a subset of population that
represents the whole group

Sampling
119
Population
the group of people we wish to
understand. Populations are often
segmented by demographic or
psychographic features (age, gender,
interests, lifestyles, etc.)
Sample
a subset of population that
represents the whole groupRespondents
people who answer
Most research cannot test everyone. Instead a sample of the
whole population is selected and tested.
If this is done well, the results can be applied to the whole
population.
This selection and testing of a sample is called sampling.
If a sample is poorly chosen, all the data may be useless.

Sampling: Two General Methods
120
Image By Sergio Valle Duarte (Own work) [CC BY 3.0], via Wikimedia Commons

Sampling Techniques
Convenience
Sampling
Judgmental
Sampling
Quota
Sampling
Snowball
Sampling
Non-probability Probability
Simple Random
Sampling
Systematic
Sampling
Stratified
Sampling
Cluster
Sampling
Other Sampling
Techniques
Proportionate Disproportionate

4.Sampling
4.4 Sample Size
122

Sampling Techniques
Convenience
Sampling
Judgmental
Sampling
Quota
Sampling
Snowball
Sampling
Simple Random
Sampling
Systematic
Sampling
Stratified
Sampling
Cluster
Sampling
Other Sampling
Techniques

Convenience Sampling
124
Convenience sampling attempts to obtain a sample of
convenient respondents. Often, respondents are selected
because they happen to be in the right place at right time.
• students or members of social organizations
• mall intercept interviews without qualifying the
respondents
• “people on the street” interviews
• tear-out questionnaires in magazines

Judgmental Sampling
125
Judgmental sampling a form of convenience sampling in
which the population elements are selected based on the
judgement of the researcher
• test markets
• purchase engineers selected in industrial marketing
research
• mothers as diaper “users”

Quota Sampling
126
Quota sampling techniques develop control categories, or
quotas, of population elements (e.g., sex, age, race, income,
company size, turnover, etc.) so that the proportion of the
elements possessing these characteristics in the sample
reflects their distribution in the population.
The elements themselves are selected based on convenience
or judgment. The only requirement, however, is that the
elements selected fit the control characteristics (quota).
Control
Characteristic
Population
Composition Sample Composition
Percentage Percentage Number
Sex
Male 
Female
 
48
52
-------
100
48 
52 
-------
100
 
480 
520 
-------
1000
Age 
18-30
31-45
45-60 
Over 60
27
39
16
18
-------
100
27
39
16
18
-------
100
270
390
160
180
-------
1000

Snowball Sampling
127127
An initial group of respondents is selected (usually) at
random.
• After being interviewed, these respondents are asked to
identify others who belong to the target population of
interest.
• Subsequent respondents are selected based on the
referrals.
Good for locating the desired characteristic in the population:
• reaching hard-to-reach respondents (e.g., government
services, “food stamps”, drug users)
• estimating characteristics that are rare in the population
• identifying buyer-seller pairs in industrial research

4.Sampling
4.4 Sample Size
128

Sampling Techniques
Convenience
Sampling
Judgmental
Sampling
Quota
Sampling
Snowball
Sampling
Simple Random
Sampling
Systematic
Sampling
Stratified
Sampling
Cluster
Sampling
Other Sampling
Techniques

Simple Random Sampling & Systematic Sampling
130
Systematic Sampling
• The sample is chosen by selecting a random starting point
and then picking every 𝑖-th element in succession from
the sampling frame
• The sampling interval, 𝑖, is determined by dividing the
population size 𝑁 by the sample size 𝑛, i.e., 𝑖 = 𝑁/𝑛
Simple Random Sampling
• Each element in the population has a known and equal
probability of selection
• Each possible sample of a given size (𝑛) has a known
probability of being the sample actually selected
• This implies that every element is selected independently
of every other element.
start here
select randomly
i
i
i
take every
i-th element

Stratified Sampling
131131
Stratified sampling is obtained by separating the population
into non-overlapping groups called strata and then obtaining
a proportional simple random sample from each group. The
individuals within each group should be similar in some way.
Good for:
• highlighting a specific subgroup within the population
• observing existing relationships between two or more
subgroups
• representative sampling of even the smallest and most
inaccessible subgroups in the population
• a higher statistical precision
Stratum A B C
Population Size 100 200 300
Sampling Fraction 1/2 1/2 1/2
Final Sample Size 50 100 150
Stratum A B C
Population Size 100 200 300
Sampling Fraction 1/5 1/2 1/3
Final Sample Size 20 100 100
Proportionate
Disproportionate

Cluster Sampling
132132
Cluster sampling the target population is first divided into
mutually exclusive and collectively exhaustive subpopulations,
or clusters. Than a random sample of clusters is selected,
based on SRS.
Good for:
• covering large geographic areas
• reducing survey costs
• when constructing a complete list of population elements
is difficult
• when the population concentrated in natural clusters
(e.g., blocks, cities, schools, hospitals, boxes, etc.)
For each cluster, either all the elements
are included in the sample (one-stage) or
a sample of elements is drawn
probabilistically (two-sage).

4.Sampling
4.4 Sample Size
133

Strengths and Weaknesses of Basic Sampling Techniques
134
Technique Strengths Weaknesses
Non-probability Sampling
Convenience sampling Least expensive, least time consuming, most
convenient
Selection bias, sample not representative, not
recommended for descriptive or causal research
Judgmental sampling Low cost, convenient, not time consuming Does not allow generalization, subjective
Quota sampling Sample can be controlled for certain characteristics Selection bias, no assurance of representativeness
Snowball sampling Can estimate rare characteristics Time consuming in the field research
Probability Sampling
Simple random sampling (SRS) Easily understood, results projectable Difficult to construct sampling frame, expensive, lower
precision, no assurance of representativeness
Systematic sampling Can increase representativeness, easier to implement
than SRS
Can decrease representativeness
Stratified sampling Includes all important subpopulations, precision Difficult to select relevant stratification variables, not
feasible to stratify on many variables, expensive
Cluster sampling Easy to implement, cost effective Imprecise, difficult to compute and interpret results

4.Sampling
4.4 Sample Size
135

Determining the Sample Size
136
The sample size does not depend on the size of the
population being studied, but rather it depends on qualitative
factors of the research.
• desired precision of estimates
• knowledge of population parameters
• number of variables
• nature of the analysis
• importance of the decision
• incidence and completion rates
• resource constraints

Sample Sizes Used in Marketing Research Studies
137
Type of Study Minimum Size Typical Size
Problem identification research
(e.g., market potential)
500 1,000 - 2,000
Problem solving research
(e.g., pricing)
200 300 - 500
Product tests 200 300 - 500
Test-market studies 200 300 - 500
TV/Radio/Print advertising
(per commercial ad tested)
150 200 - 300
Test-market audits 10 stores 10 - 20 stores
Focus groups 6 groups 10 - 15 groups

Margin of Error Approach to Determining Sample Size
138

139

140

141
𝑥 = 𝑥( ± 𝐸
𝑥 = real population parameter
𝑥( = sample statistic
𝐸 = margin of error
𝐸 = 𝑧
𝜎
𝑛

142
𝑥 = 𝑥( ± 𝐸
𝐸 = 𝑧
𝜎
𝑛
unlikely to be known

143
𝑥 = 𝑥( ± 𝐸
𝐸 = 𝑧
𝜎
𝑛
unlikely to be known
has a maximum at π = .5

144
𝑥 = 𝑥( ± 𝐸

145
𝑥 = 𝑥( ± 𝐸
calculations are approximate values for 95% level of confidence

146
𝐸 ≈
1
𝑛
⟹ 𝑛 ≈
1
𝐸
1

147

148
𝑛2344 = corrected sample size
𝑛 = sample size
𝑁 = size of population

𝑛2344 =
𝑛
(1 + 𝑛 − 1 / 𝑁)
149
Margin of Error 1%

150
𝑛2344 =
𝑛
(1 + 𝑛 − 1 / 𝑁)
Margin of Error 5%

151
𝑛2344 =
𝑛
(1 + 𝑛 − 1 / 𝑁)
Margin of Error 10%

A Note on Confidence Interval
152
Confidence Interval & Level of Confidence
A confidence interval estimate is an interval of numbers, along with a
measure of the likelihood that the interval contains the unknown
parameter.
The level of confidence is the expected proportion of intervals that will
contain the parameter if a large number of samples is maintained.
.
Suppose we're wondering what the average number of hours that people at
Siemens spend working. We might take a sample of 30 individuals and find a sample
mean of 7.5 hours. If we say that we're 95% confident that the real mean is
somewhere between 7.2 and 7.8, we're saying that if we were to repeat this with
new samples, and gave a margin of ±0.3 hours every time, our interval would
contain the actual mean 95% of the time.

Confidence Interval, Margin of Error, and Sample Size
153
The higher the confidence we need, the wider
the confidence interval and the greater the
margin of error will be

Confidence Interval, Margin of Error, and Sample Size
154
The higher the confidence we need, the wider
the confidence interval and the greater the
margin of error will be
smaller margins of error
require larger samples
higher levels of confidence
require larger samples

5.Data Analysis: A Concise Overview of Statistical Techniques
5.1 Descriptive Statistics: Some popular Displays of Data
5.2 Inferential Statistics: Can the Results Be Generalized to Population?
155

Types of Statistical Data Analysis
156
Descriptive
• Descriptive statistics provide simple
summaries about the sample and about the
observations that have been made.
• Include the numbers, tables, charts, and
graphs used to describe, organize, summarize,
and present raw data.
Inferential
• Inferential statistics are techniques that allow
making generalizations about a population
based on random samples drawn from the
population.
• Allow assessing causality and quantifying
relationships between variables.

157

158

blue red blue orange blue yellow green red pink
blue green blue purple blue blue green yellow pink
blue red pink green blue yellow green blue
Frequency and Relative Frequency Tables
159
Original Data
𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 =
𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦
𝑠𝑢𝑚 𝑜𝑓 𝑎𝑙𝑙 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑖𝑒𝑠
A frequency distribution lists each category
of data and the number of occurrences for
each category
The relative frequency is the proportion (or percent)
of observations within a category.
A relative frequency distribution lists each category
of data together with the relative frequency of each
category.
favorite color frequency
blue 10
red 3
orange 1
yellow 3
green 5
pink 3
purple 1
favorite color relative frequency
blue 10/26≈ 0.38
red 3/26≈ 0.12
orange 1/26≈ 0.04
yellow 3/26≈ 0.12
green 5/26≈ 0.19
pink 3/26≈ 0.12
purple 1/26≈ 0.04

favorite color relative frequency
blue 10/26≈ 0.38
red 3/26≈ 0.12
orange 1/26≈ 0.04
yellow 3/26≈ 0.12
green 5/26≈ 0.19
pink 3/26≈ 0.12
purple 1/26≈ 0.04
favorite color frequency
blue 10
red 3
orange 1
yellow 3
green 5
pink 3
purple 1
Bar Graphs
160
0
2
4
6
8
10
12
blue red orange yellow green pink purple
FREQUENCY
favorite color
0%
5%
10%
15%
20%
25%
30%
35%
40%
blue red orange yellow green pink purple
RELATIVE FREQUENCY
favorite color
Bar Graphs / Bar Charts
1. heights can be frequency
or relative frequency
2. bars must not touch

Pie Charts
161
blue
37%
red
12%orange
4%
yellow
12%
green
19%
pink
12%
purple
4%
favorite color
Pie Charts
1. should always include the relative
frequency
2. also should include labels, either directly
or as a legend

162

Exam Score Frequency
50–59 2
60–69 5
70–79 7
80–89 7
90–99 4
children frequency Relative frequency
1 3 3/26≈0.12
2 8 8/26≈0.31
3 10 10/26≈0.38
4 2 2/26≈0.08
5 3 3/26≈0.12
Tables
163
Original Data
Original Data
Sometimes there are too many values to
make a row for each one. In that case, we'll
need to group several values together.
A discrete variable is a quantitative variable
that has either a finite number of possible
values or a countable number of values, i.e.,
0, 1, 2, 3, ...
2 2 2 4 5 3 3 3 3
2 1 2 3 5 3 4 3 1
2 3 5 3 2 1 3 2
62 87 67 58 95 94 91 69 52
76 82 85 91 60 77 72 83 79
63 88 79 88 70 75 75
lower class limit
upper class limit
class width= 90-80 = 10

average
commute frequency
relative
frequency
16–17.9 1 1/15≈0.07
18–19.9 2 2/15≈0.13
20–21.9 1 1/15≈0.07
22–23.9 6 6/15≈0.40
24–25.9 2 2/15≈0.13
26–27.9 1 1/15≈0.07
28–29.9 1 1/15≈0.07
30–31.9 1 1/15≈0.07
children frequency
relative
frequency
1 3 3/26≈0.12
2 8 8/26≈0.31
3 10 10/26≈0.38
4 2 2/26≈0.08
5 3 3/26≈0.12
Tables
164
0
2
4
6
8
10
12
1 2 3 4 5
FREQUENCY
NUMBER OF CHILDREN IN FAMILY
0,00
0,10
0,20
0,30
0,40
0,50
1 2 3 4 5
RELATIVE FREQUENCY
0
1
2
3
4
5
6
7
16 18 20 22 24 26 28 30 32
FREQUENCY
TIME (MINUTES)
Average Daily Commute

Histogram
1. height of rectangles is the frequency or
relative frequency of the class
2. widths of rectangles is the same and
they touch each other
0
2
4
6
8
10
12
1 2 3 4 5
FREQUENCY
0,00
0,10
0,20
0,30
0,40
0,50
1 2 3 4 5
RELATIVE FREQUENCY
0
1
2
3
4
5
6
7
16 18 20 22 24 26 28 30 32
FREQUENCY
TIME (MINUTES)
Histogram
165
average
commute frequency
relative
frequency
16–17.9 1 1/15≈0.07
18–19.9 2 2/15≈0.13
20–21.9 1 1/15≈0.07
22–23.9 6 6/15≈0.40
24–25.9 2 2/15≈0.13
26–27.9 1 1/15≈0.07
28–29.9 1 1/15≈0.07
30–31.9 1 1/15≈0.07

Frequency Polygon
166
0
1
2
3
4
5
6
7
16 18 20 22 24 26 28 30 32
FREQUENCY
TIME (MINUTES)
A frequency polygon
is drawn by plotting a point above each class
midpoint and connecting the points with a
straight line.
(Class midpoints are found by average
successive lower class limits.)
16 21 26 31
0
1
2
3
4
5
6
7
16 18 20 22 24 26 28 30 32
FREQUENCY
TIME (MINUTES)
0
1
2
3
4
5
6
7
15 17 19 21 23 25 27 29 31 33
FREQUENCY
TIME (MINUTES)
average
commute frequency
relative
frequency
16–17.9 1 1/15≈0.07
18–19.9 2 2/15≈0.13
20–21.9 1 1/15≈0.07
22–23.9 6 6/15≈0.40
24–25.9 2 2/15≈0.13
26–27.9 1 1/15≈0.07
28–29.9 1 1/15≈0.07
30–31.9 1 1/15≈0.07

Cumulative Tables and Ogives
167
average
commute
relative
frequency
cumulative
relative
frequency
16–17.9 1/15≈ 0.07 1/15≈ 0.07
18–19.9 2/15≈ 0.13 2/15≈ 0.20
20–21.9 1/15≈ 0.07 1/15≈ 0.27
22–23.9 6/15≈ 0.40 6/15≈ 0.67
24–25.9 2/15≈ 0.13 2/15≈ 0.80
26–27.9 1/15≈ 0.07 1/15≈ 0.87
28–29.9 1/15≈ 0.07 1/15≈ 0.94
30–31.9 1/15≈ 0.07 1/15≈ 1.00
Cumulative tables
show the sum of values up to and including
that particular category.
An ogive
is a graph that represents the cumulative
frequency or cumulative relative frequency
for the class.
average
commute frequency
cumulative
frequency
16–17.9 1 1
18–19.9 2 3
20–21.9 1 4
22–23.9 6 10
24–25.9 2 12
26–27.9 1 13
28–29.9 1 14
30–31.9 1 15
0
0,2
0,4
0,6
0,8
1
1,2
17 19 21 23 25 27 29 31 33
Cumulative Relative Frequency
Time (minutes)

168

Measures of Central Tendency
169
Mean
𝑥̅ =
𝑥H + 𝑥1 + ⋯ + 𝑥J
𝑛
=
∑ 𝑥L
𝑛 Sum of each item Sum of average items
Mean is the “center of gravity” -
like the balance point
Advantages:
• It works well for lists that are simply combined (added)
together.
• Easy to calculate: just add and divide.
• It’s intuitive — it’s the number “in the middle”, pulled up by
large values and brought down by smaller ones.
Disadvantages:
• The average can be skewed by outliers — it doesn’t deal
well with wildly varying samples.
• The average of 100, 200 and -300 is 0, which is misleading.

170
Median
Median is the item in the middle
of a sorted list
Advantages:
• Handles outliers well — often the most accurate
representation of a group
• Splits data into two groups, each with the same number of
items
Disadvantages:
• Can be harder to calculate: you need to sort the list first
• Not as well-known; when you say “median”, people may
think you mean “average”
50% below 50% above
𝑥M = N
𝑥(OPH)/1
1
2
𝑥O/1 + 𝑥O/1PH
for odd n
for even n

171
Mode
count
item
Mode is the most frequent
observation of the variable
Advantages:
• Works well for exclusive voting situations (this choice or
that one; no compromise), i.e., for nominal data
• Gives a choice that the most people wanted (whereas the
average can give a choice that nobody wanted).
• Simple to understand
Disadvantages:
• Requires more effort to compute (have to tally up the votes)
• “Winner takes all” — there’s no middle path
The mode of
is

Measures of Central Tendency:
Using Mean and Median to Identify the Distribution Shape
172
symmetric
mean and median
approximately equal
left-skewed
median
mean is
“pulled” down
right-skewed
median
mean is
“pulled” up

Measures of Dispersion
173
𝜎1
=
∑ 𝑥L − 𝜇 1
𝑛
Population
Variance
Sample
Variance 𝑠1
=
∑ 𝑥L − 𝑥̅ 1
𝑛 − 1
Variance is the average of the
squared distance form the mean
Heights of the 2008 US Men's Olympic Basketball Team

Mean acts as a balancing point. Hence, the average difference from
the mean will equal zero.
When calculating variance, all differences are squared, so that
negative differences do not compensate positive differences.
174
Sample
Variance 𝑠1
=
∑ 𝑥L − 𝑥̅ 1
𝑛 − 1
Heights of the 2008 US Men's Olympic Basketball Team
𝑥̅ =
1.5 + 2.5 + 3.5 − 0.5 + 4.5 + 1.5 − 2.5 − 6.5 + 2.5 − 0.5 − 2.5 − 3.5
12
= 0
𝑠1
=
117
12 − 1
≈ 10.6
Why Variance?

Which data set has a higher standard deviation?
175
Standard
Deviation
𝑠 = 𝑠1
Standard Deviation
keeps the units of the original measure
𝜎 = 𝜎1
𝑠 = 10,6 ≈ 3.3
𝑠1
=
117
12 − 1
≈ 10.6 square inches
inches

Relationship between the Standard Deviation and the
Shape of the Normal Distribution
176
99,7% of the data are within
3 standard deviations of the mean
95% within
2 standard deviations
68% within
1 standard
deviation
© Dan Kernler

177

Cross-Tabulations
178
Cross-Tabulations
Cross-Tabulations are tables that reflect the joint distribution of two (or
more) variables with a limited number of categories or distinct values.
• help to understand how one variable (e.g., brand loyalty) relates to
another variable (e.g., sex)
• a cross-tabulation table contains a cell for every combination of
categories of two (or more) variables
Examples:
• How many brand-loyal users are males?
• Is product use (heavy users, medium users, light
users, and non-users) related to outdoor
activities (high, medium and low)?
• Is familiarity with a new product related to age
and education levels?
• Is product ownership related to income (height,
medium, and low)?

Cross-Tabulation
179
Education
Own Expensive Automobile College Degree No College Degree
yes 32 % 21 %
no 68 % 79 %
Column total 100 % 100 %
Number of cases 250 750
Does education influence ownership of expensive automobiles?
Ownership of Expensive Automobiles by Education Level

Cross-Tabulation
180
Sometimes introducing a third variable can
reveal
spurious relationship
suppressed association
no change in initial relationship

Cross-Tabulation
181
Does education influence ownership of expensive automobiles?
Ownership of Expensive Automobiles by Education and Income Levels
Low Income High Income
Own Expensive Automobile College Degree No College Degree College Degree No College Degree
yes 20 % 20 % 40 % 40 %
no 80 % 80 % 60 % 60 %
Column total 100 % 100 % 100 % 100 %
Number of cases 100 700 150 50
Does it?

Cross-Tabulation
182
Does age influence desire to travel?
Desire to Travel Abroad by Age
Ages
Desire to travel abroad Less than 45 45 or more
yes 50 % 50 %
no 50 % 50 %
Male Female
Desire to travel abroad < 45 ≥ 45 < 45 ≥ 45
yes 60 % 40 % 35 % 65 %
no 40 % 60 % 65 % 35 %
Column total 100 % 100 % 100 % 100 %
Desire to Travel Abroad by Age and Sex

Cross-Tabulation
183
Does family size influence frequency of eating in fast-food restaurants?
Eating Frequency in Fast-Food Restaurants by Family Size
Eat frequently in fast-food
restaurants
Family size
Small Large
yes 50 % 50 %
no 50 % 50 %
Eat frequently in fast-food
restaurants
Low income High income
Small Large Small Large
yes 50 % 50 % 50 % 50 %
no 50 % 50 % 50 % 50 %
Column total 100 % 100 % 100 % 100 %
Eating Frequency in Fast-Food Restaurants by Family Size and Income

184

185

Hypothesis Testing
186
Hypothesis Testing
Hypothesis Testing is a five-step procedure using sample evidence and
probability theory to determine whether the hypothesis is a reasonable
statement.
In other words, it is a method to prove whether or not the results
obtained on a randomly drawn sample are projectable to the whole
population.
Procedure:
1. State null and alternative hypothesis
2. Select a level of significance
3. Identify the test statistic
4. Formulate a decision rule
5. Take a sample, arrive at a decision
"People are 'erroneously confident' in their knowledge and underestimate
the odds that their information or beliefs will be proved wrong. They tend
to seek additional information in ways that confirm what they already
believed."
Max Bazerman

Hypothesis Testing
187
Sex
Internet usage Male Female Row total
light 5 10 15
heavy 10 5 15
Column total 15 15 n=30
Sex and Internet Usage
Based on this sample:
Q: Are there really more heavy internet users among males
than among females in the general population?

Hypothesis Testing
188
Step 1: State null and alternative hypothesis
A null hypothesis ( 𝑯 𝟎) is a statement of status quo,
one of no difference or no effect.
An alternative hypothesis ( 𝑯 𝟏) is one in which some
difference or effect is expected.
𝑯 𝟎: There is no difference between males and females w.r.t.
internet usage.
𝑯 𝟏: Males and females expose different internet usage
behavior.
𝐼𝑈` = 𝐼𝑈a
𝐼𝑈` ≠ 𝐼𝑈a

Hypothesis Testing
189
Step 2: Select a level of significance
Significance ( 𝜶) – probability of rejecting a true null
hypothesis.
𝜷 – probability of accepting a false null hypothesis.
Null hypothesis (𝐻0)
is true
is false
Reject null hypothesis
Type I error
False positive
Correct outcome
True positive
Fail to reject null
hypothesis
Correct outcome
True negative
Type II error
False negative
𝛽
(1 − 𝛽) – power of test
𝛼 – significance

is true
is false
Type I error
False positive
Correct outcome
True positive
Fail to reject null
hypothesis
Correct outcome
True negative
Type II error
False negative
Hypothesis Testing
190
acquit a criminal
convict an innocent
Analogy: innocence in a criminal trial
𝐻0: the defendant is innocent
hypothesis.

is true
is false
Type I error
False positive
Correct outcome
True positive
Fail to reject null
hypothesis
Correct outcome
True negative
Type II error
False negative
Hypothesis Testing
191
you continue your business near
the bush but a lion is there
there is no lion but you run away
Analogy: Rustle in the bush – is it a lion?
𝐻0: there is no lion in the bush
hypothesis.

Hypothesis Testing
192
Levels of significance in marketing research
𝛼 – level of significance (1 − 𝛼) – level of confidence
0.01 (1%)
0.05 (5%)
0.99 (99%)
0.95 (95%)
hypothesis.

Hypothesis Testing
193
Step 3: Identify the test statistic
Sample Application Level of scaling Test/Comments
One Sample
Distributions Non-metric
Kolmogorow-Smirnow and χ2
test for goodness of fit; Runs test for randomness;
Binomial test for goodness of fit of dichotomous variables
Means Metric
t test, if variance is unknown
z test, if variance is known
Proportions Metric z test
Two Independent
Samples
Kolmogorow-Smirnow two-sample test for equality of two distributions
Means Metric
Two-group t test
F test for equality of variances
Proportions Metric Non-metric
z test
χ2
test
Ranking/Medians Non-metric Mann-Whitney U test is more powerful than the median test
Paired Samples
Means Metric paired t test
Proportions Non-metric
McNemar test for binary variables,
χ2
test
Ranking/Medians Non-metric Wilcoxon matched-pairs ranked-signs test is more powerful than the sign test

Hypothesis Testing
194
Sample Application Level of scaling Test/Comments
One Sample
Kolmogorow-Smirnow and χ2
test for goodness of fit; Runs test for randomness;
Binomial test for goodness of fit of dichotomous variables
Means Metric
t test, if variance is unknown
z test, if variance is known
Proportions Metric z test
Two Independent
Samples
K-S two-sample test for equality of two distributions
Means Metric
Two-group t test
F test for equality of variances
Proportions Metric Non-metric
z test
χ2
test
Ranking/Medians Non-metric Mann-Whitney U test is more powerful than the median test
Paired Samples
Means Metric paired t test
Proportions Non-metric
McNemar test for binary variables,
χ2
test
Ranking/Medians Non-metric Wilcoxon matched-pairs ranked-signs test is more powerful than the sign test
!
In our example, we deal with one-sample distribution of a non-metric variable
(light or heavy internet usage)

Hypothesis Testing
195
χ2 (chi-square) statistic for goodness of fit is used to test the statistical
significance of the observed association in a cross-tabulation
𝐻0: There is no association between the variables
χ2 (chi-square) tests the equality of frequency distributions.
Which distributions/frequencies should we test?
𝑓 𝑒 – cell frequencies that would be expected if no association were present
between the variables
𝑓 𝑜 – actual observed cell frequencies

Hypothesis Testing
196
𝑓h =
𝑛4 𝑛2
𝑛
𝑛4 – total number in the row
𝑛2 – total number in the column
𝑛 – total sample size
𝑓hi,i
=
15 j 15
30
= 7,5 𝑓hi,k
=
15 j 15
30
= 7,5
𝑓hk,i
=
15 j 15
30
= 7,5 𝑓hk,k
=
15 j 15
30
= 7,5

Hypothesis Testing
197
In our example:
𝜒1
=
(mno.m)k
o.m
+
(Hpno.m)k
o.m
+
(Hpno.m)k
o.m
+
(mno.m)k
o.m
= 0.833 + 0.833 + 0.833 + 0.833 = 3.333
𝜒1 = r
(𝑓3 − 𝑓h)1
𝑓hall cells

Hypothesis Testing
198
Step 4: Formulate a decision rule
𝑻𝑺 𝒄𝒂𝒍 – observed value of the test statistic.
𝑻𝑺 𝒄𝒓 – critical value of the test statistic for a given
significance level.
If probability of 𝑻𝑺 𝒄𝒂𝒍 < significance level (𝜶), then reject 𝑯 𝟎.
or
If 𝑻𝑺 𝒄𝒂𝒍 > 𝑻𝑺 𝒄𝒓 , then reject 𝑯 𝟎.

Hypothesis Testing
199
Step 4: Formulate a decision rule
If probability of 𝑻𝑺 𝒄𝒂𝒍 < significance level (𝜶), then reject
𝑯 𝟎.
or
If 𝑻𝑺 𝒄𝒂𝒍 > 𝑻𝑺 𝒄𝒓 , then reject 𝑯 𝟎.
𝑑𝑓
Table of critical values of χ2 for different levels of significance 𝛼
𝑑𝑓 – degrees of freedom
𝑟 – number of rows
𝑐 – number of columns
𝑑𝑓 = 𝑟 − 1 𝑐 − 1
𝑑𝑓 = 2 − 1 2 − 1 = 1
𝜒2|}
1
= 3.333
𝜒24
1
= 3.841
3.333 < 3.841
𝜒2|}
1
< 𝜒24
1
𝐻0 cannot be rejected

Hypothesis Testing
200
Step 5: Arrive at a decision Is the evidence there?
What are the consequences?
• 𝑯 𝟎 of no association cannot be rejected
• Association is not statistically significant at the .05 level
• The findings from the sample cannot be generalized to population

Hypothesis Testing
201
Sex
Internet usage Male Female Row total
light 5 10 15
heavy 10 5 15
Column total 15 15 n=30
Sex and Internet Usage
Based on this sample:
Q: Are there really more heavy internet users among males than
among females in the general population?
A: The sample doesn’t provide such evidence.
If the sample was chosen and drawn appropriately, then we can
state that there is no such relationship in the population at the
95% confidence level.
Otherwise - we don’t know.

Principles of Survey Research (questionStar)

Principles of Survey Research (questionStar)

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (18)

Similar to Principles of Survey Research (questionStar)

Similar to Principles of Survey Research (questionStar) (20)

More from Paul Marx

More from Paul Marx (20)

Recently uploaded

Recently uploaded (20)

Principles of Survey Research (questionStar)